| Aspect | Description |
|---|---|
| Customer facing | Determine if the LLM application will directly interact with customers. Consider risks related to data privacy, accuracy, and user experience. |
| Preventing hallucinations | Assess the potential for LLMs to generate inaccurate or misleading outputs. |
| Privacy concerns & Compliance | Evaluate the privacy implications of using LLMs, particularly regarding the handling of sensitive or personally identifiable information. |
| AI bias | Consider the risk of LLMs discriminating against certain customers based on their training data. |
| Data security breach | Assess the risk of user data exposure to unauthorized parties. |
| System downtime | Evaluate the possibility of online service outages and unexpected API design changes. |
| Lack of human oversight | Consider the potential for errors or issues to go unnoticed without human oversight. |
| Misuse of AI by customers | Assess the risk of users misusing the AI for purposes that may be illegal or against terms of service. |